How to use AI in no-code, low-code, vibe coding, and coding?
Most people meet AI in a chat box and know nothing about coding. Knowing a little code makes three things easier: fixing AI’s output, running free open-source packages and models on your own machine, and mixing local and cloud tools for better value. It also makes our work with AI harder to get wrong.
Many years ago, nobody knew how to use PowerPoint, Excel, or Word. Today we expect everyone to know them. AI is following the same path.
Today, many people use AI without knowing anything about coding — this is “no-code” AI. They simply type messages into the chat interfaces of providers such as:
We can also keep typing instructions in English and ask AI to create PowerPoint, Excel, and
Word files for us. But if we also know how to make changes directly in those applications,
we can fix AI’s output quickly — or create a simple spreadsheet ourselves, adding
up numbers with a SUM formula, without asking AI at all.
Vibe coding
With “vibe coding,” we keep asking AI to do what we want simply by describing it in natural language. It is fast and forgiving for drafts, mock-ups, and one-off tasks. The limit comes when we cannot read what the AI produced and therefore cannot tell whether it is right.
Low-code: agent mode and Integrated Development Environment (IDE) mode
There are real advantages to understanding key coding concepts, or to being able to use the free programs and packages that others have already developed. Here “low-code” means a small amount of code, with AI doing most of the typing.
Hence, there is an advantage to learning how to use AI in “Agent mode” and “IDE mode” rather than relying on “chat mode” alone.
- Chat mode returns text that you copy into a file yourself.
- Agent mode edits files, runs commands, and checks results, while you review each change.
- IDE mode puts that agent inside an editor, next to the project it is changing.
The good news is that many AI companies do not want to reinvent the wheel: they use the same commands and interface as Visual Studio Code (VS Code), the free IDE from Microsoft. Most extensions, keyboard shortcuts, and settings carry over.
An analogy: no keyboard manufacturer would redesign the keyboard with a “better” layout than the current asdf layout, which our four left-hand fingers already know. Similarly, there is no need for AI companies to invent a new interface — standardizing on VS Code benefits everyone. The winning standard is not always the best one; it is the one everyone already uses.
Start here
To download and install Visual Studio Code free of charge, go to code.visualstudio.com/download. For a step-by-step setup, see my VS Code + Python setup guide with uv.
You do not need to be a developer. But being able to use Visual Studio Code, alongside PowerPoint, Excel, and Word, helps you use AI effectively — you get what you actually want — and efficiently: you can run models on your own machine and iterate there, instead of waiting for a cloud model and paying for credits on every round-trip. How many hours did you spend learning PowerPoint, Excel, and Word? Learn VS Code, and you will become an expert AI user rather than an ordinary one.
AI can then be your ally, not your opponent. In Cantonese: “AI係你對手,唔係你嘅對手。”